SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2305.07958 · IJCAI · 2023

More for Less: Safe Policy Improvement With Stronger Performance Guarantees

Marnix Suilen, Thiago Simão, Nils Jansen, Christel Baier, Patrick Wienhöft, Clemens Dubslaff

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2305.07958.json · badge: /api/ran/2305.07958.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
lava-lab/improved_spi canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
compute_Nwedge_2s Ran
ran on a synthesized input
lava-lab/improved_spi/guarantee_functions.py
code served (permissive licence) · get_code("7c7e0825f52f3a41")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from math import sqrt, log, ceil

ARGS = {
    "zeta": 0.5,
    "delta": 0.01,
    "gamma": 0.99,
    "vmax": 10.0,
    "nb_states": 50,
    "nb_actions": 4,
    "rho_b": 0.1,
    "rho_opt": 0.2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_Nwedge_beta Ran
ran on a synthesized input
lava-lab/improved_spi/guarantee_functions.py
code served (permissive licence) · get_code("42308e19a6da852c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from scipy import special as sc
from math import sqrt, log, ceil

torch.manual_seed(0)

ARGS = {
    "zeta": 5.0,
    "delta": 0.1,
    "gamma": 0.95,
    "vmax": 10.0,
    "nb_states": 4,
    "nb_actions": 3,
    "n_init": 100,
    "rho_b": 0.5,
    "rho_opt": 0.2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_zeta_spibb Ran
ran on a synthesized input
lava-lab/improved_spi/guarantee_functions.py
code served (permissive licence) · get_code("4e81fab908167675")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from math import sqrt, log, ceil

torch.manual_seed(0)

ARGS = {
    "Nwedge": 100,
    "delta": 0.05,
    "gamma": 0.99,
    "vmax": 10.0,
    "nb_states": 5,
    "nb_actions": 3,
    "rho_b": 0.2,
    "rho_opt": 0.5,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
policy_evaluation_modified Ran
ran on a synthesized input
lava-lab/improved_spi/HCPI.py
code served (permissive licence) · get_code("689e83fa371be0c2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set up small, distinct sizes for different dimensions
num_states = 4
num_actions = 3

# pi: policy, array of shape |S| x |A|
# Each row should sum to 1 (probability distribution over actions for each state)
pi = np.array([
    [0.5, 0.3, 0.2],
    [0.2, 0.5, 0.3],
    [0.3, 0.2, 0.5],
    [0.4, 0.4, 0.2]
], dtype=np.float32)

# r: rewards, array of shape |S| x |A|
r = np.array([
    [1.0, 2.0, 0.5],
    [0.5, 1.5, 2.0],
    [2.0, 0.5, 1.0],
    [1.5, 1.0, 0.5]
], dtype=np.float32)

# p: state transition probabilities, array of shape |S| x |A| x |S|
# For each state-action pair, distribution over next states
p = np.zeros((num_states, num_actions, num_states), dtype=np.float32)
for s in range(num_states):
    for a in range(num_actions):
        # Create a valid probability distribution over next states
        probs = np.random.dirichlet(np.ones(num_states))
        p[s, a, :] = probs

# gamma: discount factor, should be between 0 and 1
gamma = 0.99

# theta: convergence threshold
theta = 1e-6

ARGS = {
    'gamma': gamma,
    'pi': pi,
    'r': r,
    'p': p,
    'theta': theta
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
step Ran
ran on a synthesized input
lava-lab/improved_spi/env_util.py
code served (permissive licence) · get_code("00a69808cc4ed02d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

class SimpleEnv:
    def __init__(self):
        self.nb_states = 5
        self.terminal_states = [4]
        # transitions: shape (nb_states, nb_actions, nb_states)
        # Each entry is a probability distribution over next states
        self.transitions = np.zeros((5, 3, 5))
        for s in range(5):
            for a in range(3):
                probs = np.random.dirichlet(np.ones(5))
                self.transitions[s, a] = probs
        # rewards: shape (nb_states, nb_actions)
        self.rewards = np.random.randn(5, 3)

env = SimpleEnv()
state = 1
action = 0

ARGS = {
    "env": env,
    "state": state,
    "action": action
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fast_step Not yet run
unverified (harvested, no recorded run)
lava-lab/improved_spi/env_util.py
code served (permissive licence) · get_code("c5a02592fcea1dd2")
How it ran

no invocation recorded.

generate_batch Not yet run
unverified (harvested, no recorded run)
lava-lab/improved_spi/env_util.py
code served (permissive licence) · get_code("508a316ff79e57cd")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

In an offline reinforcement learning setting, the safe policy improvement (SPI) problem aims to improve the performance of a behavior policy according to which sample data has been generated. State-of-the-art approaches to SPI require a high number of samples to provide practical probabilistic guarantees on the improved policy's performance. We present a novel approach to the SPI problem that provides the means to require less data for such guarantees. Specifically, to prove the correctness of these guarantees, we devise implicit transformations on the data set and the underlying environment model that serve as theoretical foundations to derive tighter improvement bounds for SPI. Our empirical evaluation, using the well-established SPI with baseline bootstrapping (SPIBB) algorithm, on standard benchmarks shows that our method indeed significantly reduces the sample complexity of the SPIBB algorithm.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2305.07958")
get_code_for_paper("2305.07958")
have("2305.07958")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2305.07958.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2305.07958.svg)](https://syntology.ai/paper/2305.07958)

Connect an agent — have() is free.